Over 1,000 AI Experts Urge US to Slow Development: What Does It Mean?

What Happened?

On July 29, 2025, a group of 1,134 employees from leading artificial intelligence companies — OpenAI, Google DeepMind, Anthropic, and others — signed an open letter titled Pacing the Frontier. In it, they urge the US government to support an international effort to slow the development of the most advanced AI models, known as 'frontier' models. The letter, published on the initiative's website, does not call for a total moratorium but proposes 'pacing' mechanisms that include regulations, safety standards, and global cooperation. The signatories argue that the current speed of progress exceeds society's ability to understand and mitigate associated risks.

This event occurs in a historical context where concerns about AI safety have been growing. Since the launch of ChatGPT in 2022, the industry has seen unprecedented acceleration, with models like GPT-4, Gemini Ultra, and Claude 3 surpassing general reasoning tests. Incidents such as the spread of AI-generated disinformation in the 2024 elections, the use of models for sophisticated cyberattacks, and the lack of transparency in training data have fueled public debate. The letter represents an alarm signal from within the industry itself, unprecedented in its scale.

Why Is It Important?

This is the first time such a large group of AI insiders — people who directly work on its creation — have spoken out in a coordinated manner calling for limits. Historically, critical voices came from external academics, like philosopher Nick Bostrom, or organizations like the Future of Life Institute, which in 2023 published a letter calling for a six-month pause in training models more powerful than GPT-4. That letter was signed by figures like Elon Musk and Steve Wozniak but had few employees from leading companies. Now, it is the engineers and scientists themselves who are warning, giving the warning a different weight.

The letter is set against a backdrop of increasing global regulation. The European Union already approved the AI Act in 2024, classifying systems by risk. The United States, on the other hand, has advanced in a fragmented manner: White House executive orders, such as the one from October 2023, and legislative proposals in Congress that have yet to be consolidated. The AI Safety Summit held at Bletchley Park (UK) in November 2023 laid the groundwork for international cooperation, but without binding agreements. The July 2025 letter pressures for concrete measures before the next summit scheduled for fall in San Francisco.

"The speed at which we are advancing is incompatible with safety. We need governments to help us put emergency brakes." — Excerpt from the letter, cited by The Next Web.

The market impact is significant. Shares of AI companies could experience short-term volatility as investors interpret the letter as a regulatory risk. However, it could also benefit startups focused on safe and ethical AI, which might attract talent and funding. For business users, the signal is ambivalent: on one hand, it suggests that the creators themselves doubt the technology's maturity; on the other, it could lead to more reliable products with better safety guarantees in the long term.

What Consequences Will It Have?

The immediate impact is political: the letter pressures the US administration to accelerate its regulatory plans. The White House has already announced it will evaluate the letter's proposals, and it is expected to influence the AI Safety Summit in fall 2025. Additionally, it could foster the creation of an international body similar to the Intergovernmental Panel on Climate Change (IPCC), but for AI, as some experts have suggested.

In the business sphere, the letter reflects internal divisions. While CEOs like Sam Altman (OpenAI), Demis Hassabis (Google DeepMind), and Dario Amodei (Anthropic) have called for regulation in the past, they have never had such explicit backing from their teams. This could force companies to voluntarily adopt stricter standards to avoid a talent drain to safe AI startups, like Anthropic (founded by former OpenAI employees precisely due to safety disagreements). In fact, according to The Next Web, some signatories have stated they are considering leaving their positions if more robust safety measures are not implemented.

For users and companies adopting AI, the letter has practical implications. In the short term, debates over model transparency and public access to safety evaluations are likely to intensify. Companies may be pressured to publish more detailed 'model cards,' as Meta did with its Llama 3 model. Additionally, business clients may demand contractual clauses guaranteeing safety and independent auditing of the models they use.

What Should Readers Know?

• The letter is not binding: it is a statement of intent, not a regulation. Its effectiveness will depend on the response from the government and companies.
• The signatories represent a minority fraction of the total industry employees. It is estimated that OpenAI, Google DeepMind, and Anthropic alone employ over 10,000 workers, so 1,134 signatures represent about 10% of that base. However, the fact that they have organized publicly is significant.
• There is internal criticism: some engineers consider the letter alarmist or that it calls for control that could stifle beneficial innovation in areas like medicine, climate, and education. For example, AI models are already used for drug discovery or predicting extreme weather events, and a slowdown could delay these advances.
• Organizations like the Partnership on AI have cautiously welcomed the letter, calling for more specificity in the proposals. In a statement, they noted that "pacing must be balanced with the need to continue researching safety and social benefits."
• Compared to previous events, like the 2023 letter calling for a pause, this new initiative is more moderate in its demands but broader in its support base. It reflects a maturation of the AI safety movement, moving from calling for a total moratorium to proposing gradual control mechanisms.

In summary, Pacing the Frontier is a symptom of the industry's maturity: when the builders themselves ask for a slowdown, the public conversation about AI changes tone. The ball is now in the regulators' court, but also in the companies', which must decide whether to adopt voluntary measures or wait for legislation to force them. For readers, the lesson is clear: AI is advancing faster than our ability to govern it, and those who know it best are the first to call for caution.